- NEWS
Building robust AI systems that always outperform people might be harder than thought, say researchers who studied Go-playing bots.
Talk of superhuman artificial intelligence (AI) is heating up. But research has revealed weaknesses in one of the most successful AI systems — a bot that plays the board game Go and can beat the world’s best human players — showing that such superiority can be fragile. The study raises questions about whether more general AI systems will suffer from vulnerabilities that could compromise their safety and reliability, and even their claim to be ‘superhuman’.
doi: https://doi.org/10.1038/d41586-024-02218-7
References
Tseng, T., McLean, E., Pelrine, K., Wang, T. T. & Gleave, A. Preprint at arXiv https://doi.org/10.48550/arXiv.2406.12843 (2024).
Wang, T. T. et al. Preprint at arXiv https://doi.org/10.48550/arXiv.2211.00241 (2022).
Related Articles
-
DeepMind AI accurately forecasts weather — on a desktop computer
-
How does ChatGPT ‘think’? Psychology and neuroscience crack open AI large language models
-
ChatGPT broke the Turing test — the race is on for new ways to assess AI
-
Google AI could soon use a person’s cough to diagnose disease
Article collection: Science and the new age of AI
